US10204122B2 - Implementing an interface between tuple and message-driven control entities - Google Patents

Abstract

Some embodiments of the invention provide a novel method for interfacing between a first tuple-based controller and a second controller using a message-based protocol. The method of some embodiments identifies a set of changed tuples stored in a set of output tables, generates a set of messages based on the changed tuples, and sends the generated set of messages to a second controller. In some embodiments, the first and second controllers are parts of a network control system that manages forwarding elements to implement a logical network.

Description

BACKGROUND

There is a growing movement, driven by both industry and academia, towards a new network control paradigm called Software-Defined Networking (SDN). In Software-Defined Networking (SDN), a control plane implements and maintains the control logic that governs the forwarding behavior of shared network switching elements on a per user basis. A logical network that is implemented for a tenant of a hosting system is a good example of an SDN. The virtual (logical) network of a tenant of the hosting system connects a set of data compute nodes (e.g., virtual machines, etc.) that are assigned to the tenant, to each other and to other virtual and/or physical networks through a set of logical switches and logical routers.

In some cases, a network control system manages the control plane using multiple levels of control entities using various systems. Communicating between a tuple-based control entity and a message-driven control entity poses many difficulties as the first relies on eventual consistency to generate output from input data, while the other requires a strict protocol and state machine.

BRIEF SUMMARY

Some embodiments of the invention provide a novel method for interfacing between a first tuple-based controller and a second controller using a message-based protocol. The method of some embodiments identifies a set of changed tuples stored in a set of output tables, generates a set of messages based on the changed tuples, and sends the generated set of messages to a second controller. In some embodiments, the first and second controllers are parts of a network control system that manages forwarding elements to implement a logical network.

In some embodiments, the first controller (or local controller) manages a local control plane for a set of managed forwarding elements, modifying the forwarding behaviors for the set of managed forwarding elements according to logical definitions provided from the logical control plane. The second controller (or central controller) of some embodiments manages the logical control plane, receiving definitions of logical forwarding elements in a logical network and creating logical forwarding data for the first controller to process and implement on the managed forwarding elements of the physical network. The local controllers of some embodiments use a rules engine (e.g., nLog) to translate input data tuples to output data tuples that can be sent to the managed forwarding elements to implement the logical network.

The tuple-based local controller of some embodiments uses cacheable output tables to store processed data tuples. Messages that are generated based on the processed data tuples can be sent to the central controller at various times (e.g., after the tuple-based controller establishes a connection with the central controller(s), upon request from the central controllers, etc.). In some embodiments, the local controller uses both cacheable and non-cacheable output tables to generate messages for the central controller. For example, the local controller of some embodiments uses cacheable output tables to store physical information (e.g., virtual interface (VIF)) information related to the managed forwarding elements, and uses non-cacheable output tables for logical information (e.g., logical addresses for machines operating on the managed forwarding elements) based on data received from the central controller. In some embodiments, the cacheable tables are used for any table that stores computed state data that can be sent to the central controller. In some such embodiments, tables used for inputs or for intermediate processing are not stored in cacheable tables to improve performance of the system.

The local controller of some embodiments uses messages (or data tuples) stored in the output tables to send messages in a particular format or according to a particular protocol (e.g., protobuf) to the central controller. In some embodiments, the local controller establishes dependencies between records in the various output tables to ensure that the generated messages are sent to the central controller in a defined sequence according to the particular protocol. The dependencies between the records of the different tables in some embodiments are dependent on the type of operation that is being performed for the data tuples in the output tables. For example, in some embodiments, while a first record in a first table will depend on a second record in a second table for a first operation, the second record in the second table may depend on the first record for another operation.

The dependencies between the data tuples of the different tables may span over multiple tables and multiple levels. For example, a single data tuple may depend on multiple data tuples in multiple other tables, multiple data tuples from multiple tables may depend on a single data tuple, and a data tuple that depends on another data tuple may in turn have other data tuples that depend on it. The various dependencies are constructed to ensure that the messages based on the output data tuples are sent to the central controller in a specific order. In some embodiments, the dependencies for the data tuples of the output tables are defined based on a virtual network identifier (VNI) associated with a logical network that connects to machines coupled to the managed forwarding elements managed by the local controller.

In some embodiments, the local controller implements the dependence of data tuples in different tables by registering the tables that include the dependent data tuples for notifications regarding the particular data tuples on which the dependent data tuples depend. In some embodiments, the local controller updates the dependent data tuples with references to the particular data tuples when the notification is received. The references in some embodiments are implemented using a counter that indicates the number of remaining dependencies that must be satisfied before a message based on the dependent data tuple can be sent.

The preceding Summary is intended to serve as a brief introduction to some embodiments of the invention. It is not meant to be an introduction or overview of all of the inventive subject matter disclosed in this document. The Detailed Description that follows and the Drawings that are referred to in the Detailed Description will further describe the embodiments described in the Summary as well as other embodiments. Accordingly, to understand all the embodiments described by this document, a full review of the Summary, Detailed Description and the Drawings is needed. Moreover, the claimed subject matters are not to be limited by the illustrative details in the Summary, Detailed Description and the Drawing, but rather are to be defined by the appended claims, because the claimed subject matters can be embodied in other specific forms without departing from the spirit of the subject matters.

BRIEF DESCRIPTION OF THE DRAWINGS

The novel features of the invention are set forth in the appended claims. However, for purposes of explanation, several embodiments of the invention are set forth in the following figures.

FIG. 1 illustrates an example of a network control system that sends messages based on tuples in a set of output tables.

FIG. 2 illustrates an example of a local controller for communicating with a central controller in a network control system in order to manage a managed forwarding element.

FIG. 3 conceptually illustrates a process for generating and sending messages based on tuples in a set of output tables.

FIG. 4 illustrates an example of sending messages from a tuple-based system using cacheable and non-cacheable output tables.

FIG. 5 illustrates another example of sending messages from a tuple-based system using cacheable output tables.

FIG. 6 illustrates an example of creating dependencies between data tuples of a set of output tables.

FIG. 7 illustrates an example of sending messages for inserted tuples in a set of dependent output tables.

FIG. 8 illustrates an example of sending messages for deleted tuples of a set of dependent output tables.

FIG. 9 conceptually illustrates a process for using dependencies to generate and send messages based on tuples in a set of output tables.

FIG. 10 conceptually illustrates an electronic system with which some embodiments of the invention are implemented.

DETAILED DESCRIPTION OF THE INVENTION

In the following detailed description of the invention, numerous details, examples, and embodiments of the invention are set forth and described. However, it should be understood that the invention is not limited to the embodiments set forth and that the invention may be practiced without some of the specific details and examples discussed.

Some embodiments of the invention provide a novel method for interfacing between a first tuple-based controller and a second controller using a message-based protocol. The method of some embodiments identifies a set of changed tuples stored in a set of output tables, generates a set of messages based on the changed tuples, and sends the generated set of messages to a second controller. In some embodiments, the first and second controllers are parts of a network control system that manages forwarding elements to implement a logical network.

A logical network logically connects a set of end machines (e.g., virtual machines, physical servers, containers, etc.) and a set of physical machines (and other resources of the physical network) using a set of logical forwarding elements (e.g., logical L2 and L3 switches). This allows the physical resources of a physical network to be allocated and shared while maintaining a logical separation between the end machines of the different logical networks.

In some embodiments, the first controller (or local controller) manages a local control plane for a set of managed forwarding elements, modifying the forwarding behaviors for the set of managed forwarding elements according to logical definitions provided from the logical control plane. The second controller (or central controller) of some embodiments manages the logical control plane, receiving definitions of logical forwarding elements in a logical network and creating logical forwarding data for the first controller to process and implement on the managed forwarding elements of the physical network. The controllers of some embodiments use a rules engine (e.g., nLog) to translate input data tuples to output data tuples that can be sent to the managed forwarding elements to implement the logical network.

The tuple-based local controller of some embodiments uses cacheable output tables to store processed data tuples. Messages that are generated based on the processed data tuples can be sent to the central controller at various times (e.g., after the tuple-based controller establishes a connection with the central controller(s), upon request from the central controllers, etc.). The messaging protocol used between the local and central controllers requires a strict adherence to ordering for the messages sent by the local controller. In some embodiments, the messaging protocol allows the central controller to maintain an accurate view of the physical network and to efficiently manage the logical network.

In some embodiments, the local controller uses both cacheable and non-cacheable output tables to generate messages for the central controller. For example, the local controller of some embodiments uses cacheable output tables to store physical information (e.g., virtual interface (VIF)) information related to the managed forwarding elements, and uses non-cacheable output tables for logical information (e.g., logical addresses for machines operating on the managed forwarding elements) based on data received from the central controller. In some embodiments, the cacheable tables are used for any table that stores computed state data that can be sent to the central controller. In some such embodiments, tables used for inputs or for intermediate processing are not stored in cacheable tables to improve performance of the system.

The local controller of some embodiments uses messages (or data tuples) stored in the output tables to send messages in a particular format or according to a particular protocol (e.g., protobuf) to the central controller. In some embodiments, the local controller establishes dependencies between records in the various output tables to ensure that the generated messages are sent to the central controller in a defined sequence according to the particular protocol. The dependencies between the records of the different tables in some embodiments are dependent on the type of operation that is being performed for the data tuples in the output tables. For example, in some embodiments, while a record in a first table will depend on a record in a second table for a first operation, the record in the second table may depend on the first record for another operation.

The dependencies between the data tuples of the different tables may span over multiple tables and multiple levels. For example, a single data tuple may depend on multiple data tuples in multiple other tables, multiple data tuples from multiple tables may depend on a single data tuple, and a data tuple that depends on another data tuple may in turn have other data tuples that depend on it. The various dependencies are constructed to ensure that the messages based on the output data tuples are sent to the central controller in a specific order. In some embodiments, the dependencies for the data tuples of the output tables are defined based on a virtual network identifier (VNI) associated with a logical network that connects to machines coupled to the managed forwarding elements managed by the local controller.

In some embodiments, the local controller implements the dependence of data tuples in different tables by registering the tables that include the dependent data tuples for notifications regarding the particular data tuples on which the dependent data tuples depend. In some embodiments, the local controller updates the dependent data tuples with references to the particular data tuples when the notification is received. The references in some embodiments are implemented using a counter that indicates the number of remaining dependencies that must be satisfied before a message based on the dependent data tuple can be sent.

An overview of the process for implementing an interface between tuple and message-driven controllers is described above. Further details and examples of messaging for tuple-based controllers are described below. Specifically, Section I describes messaging from a tuple-based controller. Section II then describes examples of messaging using cacheable output tables. Section III describes examples of ordered messaging based on output table dependencies. Finally, Section IV describes an electronic system with which some embodiments of the invention are implemented.

I. Messaging for a Tuple-Based Controller

FIG. 1 illustrates an example of a network control system that sends messages based on tuples in a set of output tables in three stages101-103. This figure shows anetwork control system100 that manages a managed forwarding element MFE1 to which two virtual machines VM1-VM2 attach. Thenetwork control system100 includes acentral control plane110 and alocal control plane120, which includes input tables122, output tables124, and arules engine126.

Thenetwork control system100 of some embodiments is for implementing logical forwarding elements of a set of logical networks on physical elements of a physical network. A logical network logically connects a set of end machines (e.g., virtual machines, physical servers, containers, other resources of the physical network, etc.) using a set of logical forwarding elements (e.g., logical L2 and L3 switches). This allows the physical resources of a physical network to be allocated and shared by multiple different logical networks while maintaining a logical separation between the end machines of the different logical networks.

In some embodiments, the managed forwarding element MFE1 is a software forwarding element (e.g., a virtual switch) that operates on a hypervisor of a host machine along with the virtual machines VM1-VM2. Although only a single software managed forwarding element is shown in this figure, it should be understood that a network control system would manage several forwarding elements, which may include both hardware and software forwarding elements.

Thenetwork control system100 includes a central control plane (CCP)110, which is implemented by a cluster of central controllers in some embodiments. Thecentral control plane110 of some embodiments manages and implements logical datapath sets (LDPS) (e.g., logical switches, logical routers, etc.) of the logical networks by communicating with the local control plane (LCP)120. In some embodiments, the LCP is implemented by a set of controllers that manage the forwarding behaviors of the managed forwarding elements by generating data tuples that are sent to the managed forwarding elements. In some embodiments, the LCP is implemented by local daemons that operate on the host machines along with software managed forwarding elements and virtual machines.

In some embodiments, the LCP uses data tuples to communicate with the managed forwarding elements because data tuples allow for a simple and efficient method for distributing state to the managed forwarding elements. Tuples allow for eventual consistency and do not require frequent and prompt updates. However, CCP does not manage the state using data tuples, but rather a message format that allows CCP to function efficiently, but requires a more strict adherence to ordered messaging. In some embodiments, in order to satisfy both the CCP and the MFEs, the LCP computes data tuples for the managed forwarding elements and translates the tuples into a message format compatible with the CCP.

Thefirst stage101 shows that the input tables122 include two tuples Tuple1 andTuple2, while the output tables124 are empty. Thefirst stage101 also shows that therules engine126 processes the tuples of input tables122. In some embodiments, therules engine126 processes the tuples of the input tables122 using a series of queries and table joins to produce output tables.

In thesecond stage102, therules engine126 has generated output tuples Tuple3 andTuple4 in the output tables124. The output tuples of some embodiments include various logical and physical information that can be sent to theCCP110 and the managed forwarding element MFE1. For example, the output tuples of the output tables124 may include tuples that define the mappings of logical elements (e.g., logical ports of logical forwarding elements) to physical elements (e.g., physical ports of the MFE1)

Thethird stage103 shows that thelocal controller120 then sends the data tuple Tuple3 to the managed forwarding element MFE1. In some embodiments, the data tuple Tuple3 defines a forwarding rule and is stored in a set of forwarding tables of MFE1 to process packets for virtual machines VM1 and VM2.

Finally, thefourth stage104 shows that, in addition to sending the data tuples to MFEL the local controllers (or LCP120) sends amessage155 based on the output tuples stored in the output tables124. In some embodiments, messages from the LCP must be provided to the CCP in a specific format or in a specific order, according to a protocol (e.g., protobuf) specified for communications between theLCP120 and theCCP110.

FIG. 2 illustrates an example of a local controller that communicates with a central controller in a network control system in order to manage a managed forwarding element. This figure shows acentral controller255 and a managed forwarding element (MFE265) that communicate with thelocal controller200, similar to the example described above with reference toFIG. 1. Thelocal controller200 includes a set of input tables210, arules engine225, a set of output tables245,message generator250, apublisher260, and acompiler235. In some embodiments, thelocal controller200, managed forwarding element (MFE)265, and VMs connected to theMFE265 all operate on a single machine. In some such embodiments, several such machines (i.e., machines with a local controller, MFE, and VMs) are managed by a single instance of thecentral controller255.

Thecentral controller255 is a part of the central control plane (CCP) and converts logical datapath sets (LDPSs) into a set ofinput data tuples242 to populate the input tables210 of thelocal controller200. In some embodiments, the input tables210 include tables with logical data (e.g., access control list configurations, private virtual network configurations, port security configurations, etc.) from thecentral controller255 and with physical data (e.g., physical control plane data, virtual interface (VIF) data, etc.) from theMFEs265 managed by thelocal controller200.

Therules engine225 performs table mapping operations that convert data tuples in the input tables210 to output data tuples (e.g., messages, forwarding rules, etc.) in the output tables245. Whenever one of the input tables210 is modified, therules engine225 performs a set of table mapping operations that may result in the modification of one or more data tuples in one or more output tables. The rules engines of different embodiments detect the occurrence of an input table event differently. In some embodiments, the event processor registers for callbacks with the input tables210 for notification of changes to the data tuples of the input tables210. In such embodiments, therules engine225 performs the table mapping operations to create data tuples in the output tables245 when it receives notification from the input tables210 that data tuples have changed.

In some embodiments, thecompiler235 receivesdeclarations240 of rules and operations (e.g., table dependencies, relationships, references, etc.) to create or implement therules engine225. Thedeclarations240 describe different operations that are to be performed upon the occurrence of different events, which thecompiler235 translates into several sets of database join operations for the table mapping engine (or rules engine). In some embodiments, thedeclarations240 are used to configure the rules engine to implement dependencies between data tuples in the various tables in order to enforce a strict ordering for messages that are sent to the central controller. The dependencies are described in further detail below in Section III.

Therules engine225 maps input tables210 containing logical datapath set data and switching element attributes to the output tables245 to generate information regarding the managed switching elements. The generated information is used to send messages to the central controller and flow entries for the managed switching elements.

In some embodiments, the output tables245 store several different tables for various purposes. The output tables245 of some embodiments stores data tuples that describe both logical and physical network data. In some embodiments, the data tuples are used to communicate messages with thecentral controller255 and to communicate forwarding data to theMFE265. In some embodiments, the output tables245 include both cacheable and non-cacheable tables that are used for messaging with the central controller. Cacheable and non-cacheable tables are described in further detail below in Section II.

Thepublisher260 of some embodiments then processes the output data tuples in the output tables245 and propagates the output data tuples to the managed forwarding element(s)265. Thepublisher260 detects changes to the output tables245 and propagates the modified data tuple(s) to the managed forwardingelement265 to modify forwarding behaviors of theMFE265.

Themessage engine250 processes the output tables245 to send messages back to thecentral controller255 according to a specified communication protocol (e.g., protobuf). In some embodiments, themessage engine250 translates the data tuples of the output tables245 into the various different formats, while in other embodiments, therules engine225 is configured to generate the messages and themessage engine250 is only responsible for sending the generated messages to thecentral controller255.

FIG. 3 conceptually illustrates a process for generating and sending messages based on tuples in a set of output tables. Theprocess300 of some embodiments is performed by a local controller in a network control system. Theprocess300 begins by detecting (at305) changes in the input data tuples stored in the input tables. In some embodiments, theprocess900 detects changes by registering for notifications from the input data tables.

Theprocess300 then generates (at310) corresponding output data tuples based on the changed input data tuples. As described above, the process of some embodiments generates the corresponding output data tuples using a rules engine (or table mapping engine).

Theprocess300 then translates (at315) the generated output data tuples to messages according to a specified protocol (e.g., protobuf) for communicating with another controller (e.g., a central controller). Finally, theprocess300 sends (at320) the translated messages to the other controller (e.g., the central controller).

II. Cacheable Output Tables

FIG. 4 illustrates an example of sending messages from a tuple-based system using cacheable and non-cacheable output tables in four stages401-404. In this example, thelocal controller410 generates and sends messages from cacheable tables435 and non-cacheable output tables430. Thefirst stage401 shows alocal controller410, similar to the controllers described above with reference toFIG. 2, with input tables415, arules engine420, output tables430 and435, and amessage engine440. In addition, the output tables430 and435 include non-cacheable output tables430 and cacheable output tables435.

In thefirst stage401, thelocal controller410 has new input tuples T1 and T2 in the input table415. The input tuples T1 and T2 may be tuples that are received from the central controller, generated based on communications with managed forwarding elements, generated based on other input tables, etc. Thefirst stage401 shows that therules engine420 detects the changed tuples T1 and T2.

In thesecond stage402, therules engine420 performs table mapping and join operations generate output tuples T3 and T4 in non-cacheable output table430 and output tuples T5-T7 in cacheable output table435. In some embodiments, therules engine420 performs a series of operations on several different tables, with outputs of some operations serving as inputs for other operations. In some embodiments, the number of tuples that are generated may differ from the number of input tuples used to generate them.

In some embodiments, thelocal controller410 uses both cacheable and non-cacheable output tables to generate messages for the central controller. For example, thelocal controller410 of some embodiments uses cacheable output tables435 to store physical information (e.g., virtual interface (VIF)) information related to the managed forwarding elements, and uses non-cacheable output tables430 for logical information (e.g., logical addresses for machines operating on the managed forwarding elements) based on data received from the central controller (not shown).

Thethird stage403 shows that themessage engine440 monitors both cacheable and non-cacheable output tables430 and435 and detects changes in the output tables430 and435. Themessage engine440 of some embodiments translates the data tuples stored in the output tables430 and435 into a message format that is compatible with a specified protocol for communicating with the central controllers.

In some embodiments, the translation of data tuples to messages is not a one-to-one relationship. For example, in some embodiments, themessage engine440 concatenates and combines multiple tuples from the output tables430 and435 into a single message. Alternatively, or conjunctively, themessage engine440 of some embodiments translates a single data tuple into multiple messages for the central controller.

In some embodiments, themessage engine440 does not generate the messages at all. Rather, therules engine420 performs the table mapping and database join operations in a manner so as to generate and directly store messages for the central controller in the output tables430.

Finally, thefourth stage404 shows that themessage engine440 sends the generatedmessages450 to the central controller. The generatedmessages450 of some embodiments are used to provide physical data (e.g., VIF data) about a managed forwarding element that is managed by thelocal controller410. In some embodiments, the protocol for communications between thelocal controller410 and the central controller requires a strict ordering of the messages sent by thelocal controller410. The use of output table dependencies to enforce the ordering of the messages is described in further detail below in Section III.

Thefourth stage404 also shows that the input tables415 and the non-cacheable tables430 are empty because their contents have been processed by therules engine420 and forwarded to the central controller by themessage engine440. The cacheable output tables435, however, still have the tuples T5-T7 that were previously generated by therules engine420.

FIG. 5 illustrates another example of sending messages from a tuple-based system using cacheable output tables in three stages501-503. Thefirst stage501 shows thelocal controller410 as described inFIG. 4. In thefirst stage501, the message engine440 (or other module of the local controller410) receives arequest550 from the central controller (not shown). Therequest550 in this example is a request for all of the information that thelocal controller410 has regarding a logical forwarding element, identified by the virtual network identifier (VNI) VNI1.

In some embodiments, the central controller does not send requests for information from the local controllers. Rather the local controllers of some embodiments automatically send messages to the central controllers whenever a connection with a central controller (or multiple central controllers) is established (e.g., an initial connection, after a disconnect, etc.).

In thesecond stage502, rather than recalculating the output tuples T5-T7 from input tuples of the input tables415, themessage engine440 generates new messages from the cached data tuples T5-T7. With the rules engine of some embodiments, it is difficult to recalculate a particular set of output tuples without recreating the exact same input tuples in the input tables415. Even if the same tuples can be stored in the input tables415, reprocessing the same data can have unexpected consequences due to the relationships and interconnections of the various input and output tables.

Thethird stage503 shows that themessage engine440 sends the recreatedmessage555 to the central controller. By generating new messages from the cached output tuples of the cacheable output tables435, some embodiments of the invention avoid having to recalculate output tuples for the outgoing data messages, providing for faster responses and avoiding potential problems in the local controller.

III. Output Table Dependencies

In some embodiments, in addition to or rather than the cacheable output tables, the local controllers generate and use dependencies between the output tables to enforce a strict ordering of messages communicated to the central controllers.FIG. 6 illustrates an example of creating dependencies between data tuples of a set of output tables in four stages601-604. In some embodiments, the dependent tables and the table(s) they depend are cacheable tables, non-cacheable tables, or a mix of both.

Thefirst stage601 shows alocal controller610 with arules engine620, output tables (Tables 1-4), andmessaging engine640. Thefirst stage601 shows that tables 2 and 3 depend on table 1 (as indicated by the dashed lines). In addition, table 4 is dependent on table 3. In this figure, the dependences are shown as dashed arrows between the tables, but in some embodiments, the dependences are between data tuples of the various tables and not between the tables themselves. For example, the dependence of table 2 on table 1, may represent that in order for any tuples from table 2 to be processed by themessage engine640, a particular data tuple must first be detected in table 1. In this way, a data tuple (or multiple data tuples) from each of tables 2 and 3 is dependent on the particular data tuple in table 1. A more detailed example of the dependencies will be described in further detail below with reference toFIGS. 7 and 8.

Thefirst stage601 also shows that therules engine620 generates output data tuples in the output tables 2-4 from data in the input tables (not shown). However, as tables 2 and 3 are dependent on table 1 and table 1 has not yet received the tuple on which the data tuples of 2 and 3 depend, themessage engine640 does not generate or send any messages based on the new output data tuples. Likewise, even though table 4 depends on table 3, themessage engine640 of some embodiments will not process the tuples of table 4 until all the upstream dependencies are satisfied (i.e., until table 1 receives the required tuple).

Thesecond stage602 shows that therules engine620 has now updated table 1 with new output data tuples. In this example, as table 1 is not dependent on any other tables, themessage engine640 processes the new output data tuples of table 1 to generatemessage650 and to send it to the central controller.

In thethird stage603, dependencies for the dependent tables 2-4 have been updated with the addition of the new data tuples in table 1. In some embodiments, a first table registers for notifications from a second table when data tuples in the first table are dependent on data tuples in the second table. In such embodiments, tables 2 and 3 would have registered for notification for new tuples at table 1. When the new tuples were received instage603, table 1 would have sent notifications to the dependent tables 2 and 3, updating the dependencies, and triggeringmessage engine640 to sendmessages655 based on the tuples stored in tables 2 and 3.

Similarly, with the processing of the tuples of table 3, thefourth stage604 shows that the dependency of table 4 is satisfied, triggering themessage engine640 to send amessage660 based on the tuples of table 4 to the central controller. By configuring the tuples of table 4 to depend on the tuples of table 3, and the tuples of tables 2 and 3 to depend on the tuples of table 1, thelocal controller610 ensures that messages based on the tuples of table 1 are sent before the tuples of tables 2 and 3, and that messages based on the tuples of table 3 are sent before messages based on the tuples of table 4.

FIG. 7 illustrates an example of sending messages for inserted tuples in a set of dependent output tables in four stages701-704. In this example,local controller700 needs to join a VNI (i.e., inform the central controller that the local controller manages at least one VM or port of the VNI) and to send information about the VNI (e.g., VM information) to the central controller.

Thefirst stage701 shows alocal controller700 with amessage engine740. Thelocal controller700 also shows three tables710,720, and725. In this example, tables720 and725 are dependent on table710. The output tables710,720, and725 are currently empty, but are for storing output data tuples generated by the rules engine (not shown). In addition, each of the dependent tables720 and725 maintains a reference and a counter to indicate the dependence of the data tuples in tables720 and725 on data tuples in the table710.

In thesecond stage702, dependent output tables720 and725 have been updated with output data tuples. Dependent output table720 is for storing virtual machine (VM) information and shows a data tuple “VNI1:VM1” which indicates that VM1, attached to a managed forwarding element oflocal controller700, is a part of VNI1. Dependent output table725 stores virtual tunnel end point (VTEP) information and shows a data tuple “VNI1:VTEP1” which indicates that the address for the VTEP for VNI1 is VTEP1. However, the count (or reference) for both of these data tuples is 0, indicating that the data tuple on which they depend has not yet been received.

Table710 stores VNI information, and based on the VNI information, themessage engine740 generates messages to join the different VNIs (i.e., to inform the central controller that the local controller manages at least one VM or port of the VNI). In this case, table710 does not yet have any VNI information. In this example, the tuples in tables720 and725 are made dependent on the data tuple of table710, because the protocol rejects or ignores messages regarding a particular VNI (VNI1), when thelocal controller700 has not joined the VNI. Ifmessage engine740 sent messages from output tables720 and725 without regard to the dependencies, the messages regarding VNI1 would be ignored and could not readily be re-sent.

Thethird stage703 shows that output table710 has received a data tuple (“VNI:VNI1”) indicating that the local controller (or machines attached to the managed forwarding element of the local controller) are to join VNI1. As output table710 has no further dependencies, themessage engine710 generates amessage750 to join VNI1 and sends it to the central controller.

Finally, in thefourth stage704, the references (or counts) for dependent tables720 and725 have been updated to 1, indicating the receipt of the data tuple in table710. Themessage engine740 processes the tuples of tables720 and725 to sendmessages755 with information for VM1 and VTEP1 of VNI1 to the central controller. The necessary number of dependencies (in this case 1), can be any number of dependencies in some embodiments. As shown in this example, multiple records in multiple tables may depend on a single record. In other cases, a single data tuple may depend on multiple data tuples from one or more other tables.

The example ofFIG. 7 showed an insert operation for joining and providing data regarding a VNI. In some embodiments, the dependencies of the various tables will change based on the type of operation that is being performed on the data tuples.FIG. 8 illustrates an example of sending messages for deleted tuples of a set of dependent output tables in four stages801-804. The example of this figure is similar to the example ofFIG. 7, but rather than inserting tuples to join a VNI and to send data regarding that VNI, in this example, removes tuples to leave the VNI and remove the data regarding that VNI.

Thefirst stage801 showslocal controller800 with amessage engine840. Thelocal controller800 also shows three tables810,815, and820. In this example, table820 is dependent on tables810 and815. The output tables810 and815 are currently empty, but dependent output table820 shows an output data tuple (“VNI:VNI1”).

In this example, the output data tuple is a delete tuple (shown in italics) that indicates that the tuple is to be deleted (i.e., thelocal controller800 wants to leave VNI1). In some embodiments, each data tuple includes a flag (e.g., ‘delete=true’) that indicates whether the tuple is for insertion or deletion. Dependent table820 also maintains a counter to indicate the dependence of the data tuple on delete data tuples in tables810 and815. In this case, the count begins at 2 based on the two records that were inserted for the VNI1 in the example ofFIG. 7. As shown in this example, a data tuple that is depended upon by a second data tuple for a first operation (e.g., insert), may depend on the second data tuple for a second operation (e.g., delete).

In thesecond stage802, output data table810 has received a delete data tuple for removing the VM information. As table810 does not depend on any other tables, themessage engine840 sends message850 to the central controller based on the updated data tuple.

Aftermessage engine840 sends message850, thethird stage803 shows that the count for output data table820 has been updated to 1, indicating that only one dependency is remaining. Thethird stage803 also shows that output table815 has received a delete data tuple for removing the VTEP information and sent acorresponding message855 to the central controller.

Finally, in thefourth stage804, the count for table820 has been updated to 0, indicating that no more dependencies remain for the data tuple, allowingmessage engine840 to send amessage860 to remove thelocal controller800 from the VNI1. The use of the various dependencies and sub-dependencies (i.e., dependencies on other dependent tuples) allows for a flexible and general mechanism for enforcing a strict order for messages based on tuples generated at a controller.

FIG. 9 conceptually illustrates a process for using dependencies to generate and send messages based on tuples in a set of output tables. Theprocess900 of some embodiments is performed by a local controller in a network control system. Theprocess900 begins by identifying (at905) changed data tuples in a set of output data tables. Theprocess900 then determines (at910) whether the changed data tuples are dependent on other data tuples. When theprocess900 determines (at910) that the changed data tuples are not dependent on any other data tuples, the process continues to step920 described below.

When theprocess900 determines (at910) that the changed data tuples are dependent on other data tuples, the process determines (at915) whether the dependencies have been satisfied. As described above, theprocess900 of some embodiments determines (at915) whether dependencies have been satisfied based on counters or references that indicate the dependent relationships between the various data tuples. For example, in some embodiments, theprocess900 determines whether the value for a counter of a particular tuple matches a particular value (e.g., 0 when performing a deletion operation, some number greater than 0 when performing an insert operation). When theprocess900 determines (at915) that the dependencies have not been satisfied, the process returns to step905.

When theprocess900 determines (at915) that the dependencies have been satisfied, the process translates (at920) the generated output data tuples to messages according to a specified protocol for communicating with another controller (e.g., a central controller). Finally, theprocess900 sends (at925) the translated messages to the other controller.

IV. System

Many of the above-described features and applications are implemented as software processes that are specified as a set of instructions recorded on a computer readable storage medium (also referred to as computer readable medium). When these instructions are executed by one or more computational or processing unit(s) (e.g., one or more processors, cores of processors, or other processing units), they cause the processing unit(s) to perform the actions indicated in the instructions. Examples of computer readable media include, but are not limited to, CD-ROMs, flash drives, random access memory (RAM) chips, hard drives, erasable programmable read-only memories (EPROMs), electrically erasable programmable read-only memories (EEPROMs), etc. The computer readable media does not include carrier waves and electronic signals passing wirelessly or over wired connections.

In this specification, the term “software” is meant to include firmware residing in read-only memory or applications stored in magnetic storage, which can be read into memory for processing by a processor. Also, in some embodiments, multiple software inventions can be implemented as sub-parts of a larger program while remaining distinct software inventions. In some embodiments, multiple software inventions can also be implemented as separate programs. Finally, any combination of separate programs that together implement a software invention described here is within the scope of the invention. In some embodiments, the software programs, when installed to operate on one or more electronic systems, define one or more specific machine implementations that execute and perform the operations of the software programs.

FIG. 10 conceptually illustrates anelectronic system1000 with which some embodiments of the invention are implemented. Theelectronic system1000 may be a computer (e.g., a desktop computer, personal computer, tablet computer, etc.), server, dedicated switch, phone, PDA, or any other sort of electronic or computing device. Such an electronic system includes various types of computer readable media and interfaces for various other types of computer readable media.Electronic system1000 includes abus1005, processing unit(s)1010, asystem memory1025, a read-only memory1030, apermanent storage device1035,input devices1040, andoutput devices1045.

Thebus1005 collectively represents all system, peripheral, and chipset buses that communicatively connect the numerous internal devices of theelectronic system1000. For instance, thebus1005 communicatively connects the processing unit(s)1010 with the read-only memory1030, thesystem memory1025, and thepermanent storage device1035.

From these various memory units, the processing unit(s)1010 retrieves instructions to execute and data to process in order to execute the processes of the invention. The processing unit(s) may be a single processor or a multi-core processor in different embodiments.

The read-only-memory (ROM)1030 stores static data and instructions that are needed by the processing unit(s)1010 and other modules of the electronic system. Thepermanent storage device1035, on the other hand, is a read-and-write memory device. This device is a non-volatile memory unit that stores instructions and data even when theelectronic system1000 is off. Some embodiments of the invention use a mass-storage device (such as a magnetic or optical disk and its corresponding disk drive) as thepermanent storage device1035.

Other embodiments use a removable storage device (such as a floppy disk, flash memory device, etc., and its corresponding drive) as the permanent storage device. Like thepermanent storage device1035, thesystem memory1025 is a read-and-write memory device. However, unlikestorage device1035, thesystem memory1025 is a volatile read-and-write memory, such a random access memory. Thesystem memory1025 stores some of the instructions and data that the processor needs at runtime. In some embodiments, the invention's processes are stored in thesystem memory1025, thepermanent storage device1035, and/or the read-only memory1030. From these various memory units, the processing unit(s)1010 retrieves instructions to execute and data to process in order to execute the processes of some embodiments.

Thebus1005 also connects to the input andoutput devices1040 and1045. Theinput devices1040 enable the user to communicate information and select commands to the electronic system. Theinput devices1040 include alphanumeric keyboards and pointing devices (also called “cursor control devices”), cameras (e.g., webcams), microphones or similar devices for receiving voice commands, etc. Theoutput devices1045 display images generated by the electronic system or otherwise output data. Theoutput devices1045 include printers and display devices, such as cathode ray tubes (CRT) or liquid crystal displays (LCD), as well as speakers or similar audio output devices. Some embodiments include devices such as a touchscreen that function as both input and output devices.

Finally, as shown inFIG. 10,bus1005 also coupleselectronic system1000 to anetwork1065 through a network adapter (not shown). In this manner, the computer can be a part of a network of computers (such as a local area network (“LAN”), a wide area network (“WAN”), or an Intranet, or a network of networks, such as the Internet. Any or all components ofelectronic system1000 may be used in conjunction with the invention.

Some embodiments include electronic components, such as microprocessors, storage and memory that store computer program instructions in a machine-readable or computer-readable medium (alternatively referred to as computer-readable storage media, machine-readable media, or machine-readable storage media). Some examples of such computer-readable media include RAM, ROM, read-only compact discs (CD-ROM), recordable compact discs (CD-R), rewritable compact discs (CD-RW), read-only digital versatile discs (e.g., DVD-ROM, dual-layer DVD-ROM), a variety of recordable/rewritable DVDs (e.g., DVD-RAM, DVD-RW, DVD+RW, etc.), flash memory (e.g., SD cards, mini-SD cards, micro-SD cards, etc.), magnetic and/or solid state hard drives, read-only and recordable Blu-Ray® discs, ultra density optical discs, any other optical or magnetic media, and floppy disks. The computer-readable media may store a computer program that is executable by at least one processing unit and includes sets of instructions for performing various operations. Examples of computer programs or computer code include machine code, such as is produced by a compiler, and files including higher-level code that are executed by a computer, an electronic component, or a microprocessor using an interpreter.

While the above discussion primarily refers to microprocessor or multi-core processors that execute software, some embodiments are performed by one or more integrated circuits, such as application specific integrated circuits (ASICs) or field programmable gate arrays (FPGAs). In some embodiments, such integrated circuits execute instructions that are stored on the circuit itself. In addition, some embodiments execute software stored in programmable logic devices (PLDs), ROM, or RAM devices.

As used in this specification and any claims of this application, the terms “computer”, “server”, “processor”, and “memory” all refer to electronic or other technological devices. These terms exclude people or groups of people. For the purposes of the specification, the terms display or displaying means displaying on an electronic device. As used in this specification and any claims of this application, the terms “computer readable medium,” “computer readable media,” and “machine readable medium” are entirely restricted to tangible, physical objects that store information in a form that is readable by a computer. These terms exclude any wireless signals, wired download signals, and any other ephemeral signals.

This specification refers throughout to computational and network environments that include virtual machines (VMs). However, virtual machines are merely one example of data compute nodes (DCNs) or data compute end nodes, also referred to as addressable nodes. DCNs may include non-virtualized physical hosts, virtual machines, containers that run on top of a host operating system without the need for a hypervisor or separate operating system, and hypervisor kernel network interface modules.

VMs, in some embodiments, operate with their own guest operating systems on a host using resources of the host virtualized by virtualization software (e.g., a hypervisor, virtual machine monitor, etc.). The tenant (i.e., the owner of the VM) can choose which applications to operate on top of the guest operating system. Some containers, on the other hand, are constructs that run on top of a host operating system without the need for a hypervisor or separate guest operating system. In some embodiments, the host operating system uses name spaces to isolate the containers from each other and therefore provides operating-system level segregation of the different groups of applications that operate within different containers. This segregation is akin to the VM segregation that is offered in hypervisor-virtualized environments that virtualize system hardware, and thus can be viewed as a form of virtualization that isolates different groups of applications that operate in different containers. Such containers are more lightweight than VMs.

Hypervisor kernel network interface modules, in some embodiments, is a non-VM DCN that includes a network stack with a hypervisor kernel network interface and receive/transmit threads. One example of a hypervisor kernel network interface module is the vmknic module that is part of the ESXi™ hypervisor of VMware, Inc.

It should be understood that while the specification refers to VMs, the examples given could be any type of DCNs, including physical hosts, VMs, non-VM containers, and hypervisor kernel network interface modules. In fact, the example networks could include combinations of different types of DCNs in some embodiments.

The term “packet” is used throughout this application to refer to a collection of bits in a particular format sent across a network. It should be understood that the term “packet” may be used herein to refer to various formatted collections of bits that may be sent across a network. A few examples of such formatted collections of bits are Ethernet frames, TCP segments, UDP datagrams, IP packets, etc.

While the invention has been described with reference to numerous specific details, one of ordinary skill in the art will recognize that the invention can be embodied in other specific forms without departing from the spirit of the invention. Thus, one of ordinary skill in the art would understand that the invention is not to be limited by the foregoing illustrative details, but rather is to be defined by the appended claims.

Claims (20)

We claim:

1. A non-transitory machine readable medium storing a local first controller executing on a set of processing units of a host computer in a network managed by a central second controller, the local first controller comprising an input first set of tables, an output second set of tables, and sets of instructions for:

determining that a first set of data tuples stored in the input first set of tables has been modified;

based on the modified first set of data tuples, generating a second set of data tuples stored in the output second set of tables, the generated second set of data tuples comprising data for performing forwarding operations by a managed forwarding element executing on the host computer and managed by the local first controller to forward data messages to other managed forwarding elements in the network;

based on the generated second set of data tuples, generating a set of control messages; and

sending the generated set of control messages to the central second controller, which based on the set of control messages, generates at least one control message to distribute to at least one other local third controller to use in managing another managed forwarding element on another host computer.

2. The non-transitory machine readable medium ofclaim 1, wherein the set of instructions for determining that the first set of data tuples has been modified comprises a set of instructions for monitoring the input first set of tables.

3. The non-transitory machine readable medium ofclaim 1, wherein the set of instructions for generating the set of control messages comprises a set of instructions for:

identifying dependencies between data tuples in the second set; and

based on the identified dependencies, identifying a particular order for the control messages in the set of control messages.

4. The non-transitory machine readable medium ofclaim 1, wherein the set of instructions for generating the set of control messages comprises a set of instructions for generating a first number of control messages based on a second number of data tuples, wherein the first number and the second number are different.

5. The non-transitory machine readable medium ofclaim 1, wherein the output second set of tables comprises a subset of cacheable output tables and also comprises a subset of non-cacheable output tables, wherein the set of instructions for generating the second set of tuples comprises a set of instructions for generating a first subset of the second set of tuples stored in the subset of cacheable output tables and a second subset of the second set of tuples stored in the subset of non-cacheable output tables.

6. The non-transitory machine readable medium ofclaim 5, wherein the set of instructions for generating the first and second subsets of tuples comprises a set of instructions for performing a series of table joins on the input first set of tables to create records in the subsets of cacheable and non-cacheable output tables.

7. The non-transitory machine readable medium ofclaim 5, wherein the local first controller further comprises a set of instructions for, upon determining that a connection with the central second controller has been re-established, sending the set of control messages to the central second controller based on the first subset of the second set of tuples stored in the set of cacheable output tables.

8. The non-transitory machine readable medium ofclaim 5, wherein the set of instructions for sending the set of control messages comprises sets of instructions for:

sending a first subset of control messages based on the first subset of tuples and a second subset of control messages based on the second subset of tuples; and

removing only the second subset of tuples from the set of non-cacheable output tables.

9. The non-transitory machine readable medium ofclaim 5, wherein the subset of cacheable output tables are for storing physical information identified by the local first controller, wherein the physical information comprises virtual interface (VIF) information for the set of managed forwarding elements.

10. The non-transitory machine readable medium ofclaim 9, wherein the subset of non-cacheable output tables are for storing logical information regarding the set of managed forwarding elements, wherein the logical information comprises mappings of the VIF information to elements of a logical network that is (1) implemented by at least the managed forwarding elements and (2) managed by at least the central second controller.

11. The non-transitory machine readable medium ofclaim 10, wherein the logical information further comprises at least one of a virtual machine (VM) Internet Protocol (IP) address, a VM Media Access Control (MAC) address, and a virtual tunnel end point (VTEP) IP address.

12. The non-transitory machine readable medium ofclaim 5, wherein the set of instructions for sending the generated set of control messages comprises a set of instructions for (1) receiving a request for information from the central second controller and (2) sending the set of control messages comprising the requested information in response to the request, wherein a key is used to filter the requested information, wherein the key is a virtual network identifier (VNI) for a logical network associated with at least one machine attached to a forwarding element managed by the local first controller.

13. The non-transitory machine readable medium ofclaim 1, wherein the set of instructions for generating the set of control messages comprises a set of instructions for identifying a dependence of a first subset of data tuples stored in a first table in the output second set of tables on a second subset of data tuples stored in a second table in the output second set of tables.

14. The non-transitory machine readable medium ofclaim 13, wherein the dependence of the first subset of data tuples on the second subset of data tuples is based on a key stored in the first and second tables, wherein the key is a virtual network identifier (VNI) for a logical network associated with at least one machine attached to a forwarding element managed by the first controller.

15. The non-transitory machine readable medium ofclaim 13, wherein the set of instructions for generating the set of control messages further comprises a set of instructions for identifying a dependence of a third subset of data tuples stored in a third table in the output second set of tables on the second subset of data tuples.

16. The non-transitory machine readable medium ofclaim 13, wherein the set of instructions for generating the set of control messages further comprises a set of instructions for identifying a dependence of a third subset of data tuples stored in a third table in the output second set of tables on the first subset of data tuples.

17. The non-transitory machine readable medium ofclaim 13, wherein the second set of data tuples indicate a particular type of operation, wherein the set of instructions for generating the set of control messages comprises a set of instructions for (i) identifying a dependence of a third subset of data tuples stored in a third table in the output second set of tables on a fourth subset of data tuples stored in a fourth table in the output second set of tables when the particular type of operation is a first type of operation, and (ii) identifying a dependence of the fourth subset of data tuples on the third subset of data tuples when the particular type of operation is a second type of operation.

18. The non-transitory machine readable medium ofclaim 17, wherein the second set of data tuples comprise a flag to indicate whether the particular type of operation is an insert operation or a delete operation.

19. The non-transitory machine readable medium ofclaim 13, wherein the first table is registered for notifications for changes to the second subset of data tuples stored in the second table, wherein the set of instructions for generating the set of control messages comprises a set of instructions for updating, when the first table receives a notification regarding changes to the second subset of data tuples, a reference in the first subset of data tuples, wherein the reference indicates whether control messages based on the first subset of data tuples can be sent.

20. The non-transitory machine readable medium ofclaim 19, wherein the reference is a counter, wherein the set of instructions for sending the generated set of control messages comprises a set of instructions for determining whether the counter has a particular value.

Images